This repository holds NVIDIA-maintained utilities to streamline mixed precision and distributed training in Pytorch. Some of the code here will be included in upstream Pytorch eventually. The intent of Apex is to make up-to-date utilities available to users as quickly as possible.

Deprecated. Use PyTorch AMP

apex.amp is a tool to enable mixed precision training by changing only 3 lines of your script. Users can easily experiment with different pure and mixed precision training modes by supplying different flags to amp.initialize.

Webinar introducing Amp (The flag cast_batchnorm has been renamed to keep_batchnorm_fp32).

API Documentation

Comprehensive Imagenet example

DCGAN example coming soon...

Moving to the new Amp API (for users of the deprecated "Amp" and "FP16_Optimizer" APIs)

apex.parallel.DistributedDataParallel is deprecated. Use torch.nn.parallel.DistributedDataParallel

apex.parallel.DistributedDataParallel is a module wrapper, similar to torch.nn.parallel.DistributedDataParallel. It enables convenient multiprocess distributed training, optimized for NVIDIA's NCCL communication library.

API Documentation

Python Source

Example/Walkthrough

The Imagenet example shows use of apex.parallel.DistributedDataParallel along with apex.amp.

Deprecated. Use torch.nn.SyncBatchNorm

apex.parallel.SyncBatchNorm extends torch.nn.modules.batchnorm._BatchNorm to support synchronized BN. It allreduces stats across processes during multiprocess (DistributedDataParallel) training. Synchronous BN has been used in cases where only a small local minibatch can fit on each GPU. Allreduced stats increase the effective batch size for the BN layer to the global batch size across all processes (which, technically, is the correct formulation). Synchronous BN has been observed to improve converged accuracy in some of our research models.

To properly save and load your amp training, we introduce the amp.state_dict(), which contains all loss_scalers and their corresponding unskipped steps, as well as amp.load_state_dict() to restore these attributes.

In order to get bitwise accuracy, we recommend the following workflow:

# Initialization
opt_level = 'O1'
model, optimizer = amp.initialize(model, optimizer, opt_level=opt_level)

# Train your model
...
with amp.scale_loss(loss, optimizer) as scaled_loss:
    scaled_loss.backward()
...

# Save checkpoint
checkpoint = {
    'model': model.state_dict(),
    'optimizer': optimizer.state_dict(),
    'amp': amp.state_dict()
}
torch.save(checkpoint, 'amp_checkpoint.pt')
...

# Restore
model = ...
optimizer = ...
checkpoint = torch.load('amp_checkpoint.pt')

model, optimizer = amp.initialize(model, optimizer, opt_level=opt_level)
model.load_state_dict(checkpoint['model'])
optimizer.load_state_dict(checkpoint['optimizer'])
amp.load_state_dict(checkpoint['amp'])

# Continue training
...

Note that we recommend restoring the model using the same opt_level. Also note that we recommend calling the load_state_dict methods after amp.initialize.

Each apex.contrib module requires one or more install options other than --cpp_ext and --cuda_ext. Note that contrib modules do not necessarily support stable PyTorch releases.

NVIDIA PyTorch Containers are available on NGC: https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch. The containers come with all the custom extensions available at the moment.

See the NGC documentation for details such as:

• how to pull a container
• how to run a pulled container
• release notes

To install Apex from source, we recommend using the nightly Pytorch obtainable from https://github.com/pytorch/pytorch.

The latest stable release obtainable from https://pytorch.org should also work.

We recommend installing Ninja to make compilation faster.

For performance and full functionality, we recommend installing Apex with CUDA and C++ extensions via

git clone https://github.com/NVIDIA/apex
cd apex
# if pip >= 23.1 (ref: https://pip.pypa.io/en/stable/news/#v23-1) which supports multiple `--config-settings` with the same key... 
pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./
# otherwise
pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --global-option="--cpp_ext" --global-option="--cuda_ext" ./

APEX also supports a Python-only build via

pip install -v --disable-pip-version-check --no-build-isolation --no-cache-dir ./

A Python-only build omits:

• Fused kernels required to use apex.optimizers.FusedAdam.
• Fused kernels required to use apex.normalization.FusedLayerNorm and apex.normalization.FusedRMSNorm.
• Fused kernels that improve the performance and numerical stability of apex.parallel.SyncBatchNorm.
• Fused kernels that improve the performance of apex.parallel.DistributedDataParallel and apex.amp. DistributedDataParallel, amp, and SyncBatchNorm will still be usable, but they may be slower.

pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" . may work if you were able to build Pytorch from source on your system. A Python-only build via pip install -v --no-cache-dir . is more likely to work.
If you installed Pytorch in a Conda environment, make sure to install Apex in that same environment.

If a requirement of a module is not met, then it will not be built.